Datasets:

RobotsMali
/

afvoices

text stringlengths 3 459	duration float64 1.02 30	label-v1 stringlengths 0 332	label-v2 stringlengths 0 432
Pamori tun bɛ foo	1.34	An amadu tun bɛ ko	Famori tun bɛ ko[?]
An me kɛrɛlamana dɔw don a tigilamɔgɔ la	2.555986	An bɛ kɛlɛ lamana dɔw don a tigilamɔgɔ la	An bɛ kɛrɛla mana dɔw don a tigilamɔgɔ la
[noise] mɔgɔ caman de b'a bila u kuŋolo la	1.563991	Mɔgɔ caman de b'a bila u kunkolo la	Mɔgɔ caman de b'a bila u kunkolo la[?]
Baarakɛ yɔrɔ tɛ yen ne ni min tɛ se ka baarakɛ ɲɔgɔn fɛ.	2.3	Baarakɛ yɔrɔ ten yen min min min se ka baarakɛ ɲɔgɔn fɛ	Baarakɛ yɔrɔ tɛ yen ne ni min tɛ se ka baara kɛ ɲɔgɔn fɛ
Ɲɛ dimi bi o la[?]	1.276009	N ɲɛ den bɛ o la	Ɲɛdɛɲɛ bɛ o la
U bɛ se ka don bala cogo min na.	1.564014	U bɛ se ka don ba cogo min na	Bɛ se ka don bala cogo min na
O safinɛ bɛ bɔ yɔrɔ min na ,a sanfinɛ bɛ ja.	2.556009	O safinɛ bɛ bɔ yɔrɔ min na a sanfinɛ bɛ ja	O safunɛ bɛ bɔ yɔrɔ min na a safunɛ bɛ ja
Ni a ye paseke mɔgɔ kelen de kuŋolo be ka baara kɛ nin bɛɛ la.	3.74	Ne a ye parce mɔgɔ kelen de kunkolo bɛ ka baara kɛ ni bɛɛ la	Ni a ye paseke mɔgɔ kelen de kunkolo bɛ ka baara kɛ nin bɛɛ la
O bɛ kɛ sababu ye	1.372	O bɛ kɛ sababu ye	O bɛ kɛ sababu ye
N'i sigilen don [noise] e lafiyalen don.	1.82	N'i sigilenlen don i lafiyalen don	N'i sigilen don i lafiyalen don
Kɛ kila ka taa bɔ [?]	1.051995	Ka tila ka taa bɔ	Ka kila ka taa bɔ
Dɔw kɛmɛ fila ni bi duuru	1.66	N'a kɛmɛ fila ni bɔ duuru bɛ	Dɔw kɛmɛ fila ni bi duuru
N'olu mago ye jiba la.	1.275986	N'olu mago ye ji b'a la	N'olu mago bɛ ji ba la
N'an b'a weele koo jigini	1.884	N'an b'a wele ko jigimi	N'an b'a wele ko jigini
Se ka taa feere	1.116	Se taa foli	Se ka taa feere
U bɛ fan bɛɛ kɛ mankan ye.	1.148005	U bɛ fan bɛɛ kɛkɛ	U bɛ fan bɛɛ kɛ mankan ye
Fɛn minnu, ni fɛn minnu bɛ ka k'ala la	1.531995	Fɛn minnu ni fɛn minnu bɛ ka kala la	Fɛn minnu ni fɛn minnu bɛ ka k'a la
Bɛ ɛ sɛbɛn t'an fɛ	1.468	Sɛbɛ t'an fɛ	Bɛ ɛ sɛbɛn t'an fɛ
[cs] a kan ka kalan de [?]	1.691995	Donc a kan ka kalan de ye	[cs] a kan ka kalan de
Bɔ [cs] i be fɛn dɔ kɛ kasɔrɔ i m'a kalan kɛ [cs]	3.772	Bon binɛrɛ k i bɛ fɛn dɔ kɛ k ka sɔrɔ i ma kalan kɛ mɛ	[cs] binfɛn ke i bɛ fɛn dɔ kɛ kasɔrɔ i m'a kalan kɛ mɛ
An ga duntafɛnw b'o la [?]	1.211995	An ka duntafɛnw b' la	An ka duntanw b'o la
A bɛ se ka kɛ den in ɲɛ. A bɛ bori [?] ale yɛrɛ fɛnɛ be ko dɔw kɛ den na	2.908	A bɛ se ka den ɲɛ a bɛ g ale yɛrɛ fana bɛ kolokɛ den na	A bɛ se ka kɛ den a bɛ bɔ dɛ ale yɛrɛ fɛnɛ bɛ ko dɔ kɛ den na
A dɔw la yɛrɛ sugu kɔnɔna makan ka ca.	2.651995	A dɔw la yɛrɛ suk mankan ka ca	A dɔw la yɛrɛ sugu kɔnɔna mankan ka ca
[?] kɛ komi [?]	1.308005	Aa ko mun ye	A kɛ komi[?]
O nɔgɔya in bɔ yen	1.788005	O nɔgɔya in n bɔ yen	O nɔgɔya in bɔ yen[um]
O dɔ y'a yɛlma [noise]	1.211995	O don y'a yɛlɛma	O dɔ ye a yɛlɛma[?]
K'an bɛ taa Bamakɔ [um]	1.628005	K' bɛ taa bamakɔkɔrɔ	K'an bɛ taa bamakɔ[um]
Kɛ furanjɛ dɔɔni don u ni ɲɔgɔn cɛ [cs] u kana na se ka ɲɔgɔn deku.	2.94	Kauranjɛ dɔn don ni ɲɔgɔn cɛ pur que u kana na se ka ɲɔgɔn di	Ka furancɛ dɔɔni[cs] u kana se ka ɲɔgɔn degun
N y'a fɔ aw ɲɛna cogo min na	1.468005	N y'a fɔ' aw ɲɛna cogo min na	N y'a fɔ aw ɲɛna cogo min na[?]
Musokɔnɔmaw ta fɛnɛ ye.	1.435986	Musokɔnɔmɔgɔww ta f ye	Muso kɔnɔmaw ta fɛnɛ ye
A bi yɛlɛ n na, u b'a fɔ ko [cs] ko dɔnniya re don [?]	2.332	A bɛɛlɛnna b'a fɔ ko dɔn ko dɔnniya ye don	A bɛ yɛlɛ n na u b'a fɔ ko[cs] ko dɔnniya de don
Fɛn min kun ka d'a ye kosɛbɛ o ye gɛrɛɲɔgɔnna ye	1.883991	Fɛn min tun ka da yesɛbɛ o ye jeli ɲɔgɔnna ye	Fɛn min kun ka d'a ye kosɛbɛ o ye gɛrɛɲɔgɔnna ye
F'a ka ɲɔgɔn dɛmɛ yɛrɛ kosɛbɛ.	1.66	Fa ka ɲɔgɔn dɛmɛ yɛrɛ gosi	F'a ka ɲɔgɔn dɛmɛ yɛrɛ kosɛbɛ
E nana de ko ka	1.02	E nana de ko nka	E nana de ko ka
A kɛlɛ bolo [?]	1.18	A kɛlɛ bolo	Ka kɛlɛbolo
U kan ka kɛ	1.02	U kan ka kɛ	O kan ka kɛ[?]
Mɛ o gɛlɛya fɔlɔf ɔlɔ ye jumɛn ye?	1.372018	Mɛ o gɛlɛya fɔlɔfɔlɔ jumɛn	Mɛ o gɛlɛya fɔlɔfɔlɔ ye jumɛn
N balima bɛɛ lajɛlen ma.	1.436009	N balima bɛɛ lajɛlen ma	N balima bɛɛ lajɛlen ma
[cs] n'i ye sɛ si don dɔw yɛrɛ bɛ sɛ si don i b'a ye	2.651995	U bɛ n'i ye sɛsu don dɔw yɛrɛ bɛ sariyasu don i b'a ye	[cs] n'i ye shɛ si don dɔw yɛrɛ bɛ shɛ si don i b'a ye
[um] nowanburukalo.	1.692018	Ee n'ufangolo kalo	[um] numankuru kalo
A dɔw b'a fɔ ko a donna ta, a bɔra ta, a don yɔrɔ n'a bɔ yɔrɔ ka ca.	3.356	A dɔw b'a fɔ ko a donna ta aa bɔra ta a don yɔrɔ n'a bɔ ka ca	A dɔw b'a fɔ ko a donna ta a bɔra tan a don yɔrɔ n'a bɔ yɔrɔ ka ca
Ɔ kaba fana nana yen n na	1.788005	Ɔ kaba fana nana yen na	Ɔ kaba fana nana yen nɔ
Ne ka ɛɛ kurun [cs]	1.372	Ne ka k kuruurun sanfɛfo	Ne ka kurun[cs]
[um] an y'an t'o la ne ye kile damadamanin kɛ [?]	2.364014	A y'an t'o la ni tile damamani kɛ	[um] an y'an t'o la ne ye kile damadamanin kɛ
O cogoya la n'i y'olu lajɛ	1.883991	O cogoya la ni y'olu lajɛ	O cogoya la n'i y'olu lajɛ
Ɔ minɛn minnu bɛ yen ni [um]	2.204	Ɔ minɛ minnu bɛ yen ni	Ɔ minɛn minnu bɛ yen ni[um]
U fila tun t'u ka wari sɔrɔ.	1.436009	U fila tun du ka wari jigin	Ani fila kun bɛ to ka wari ci ne ma
Kuma ni kuma ne bɛ maakɔrɔw wele ni min kɛ ka na sigi,	2.716009	Tuma ni tuma ne bɛ maakɔrɔ wele min kɛ ka na sigi	Kuma ni kuma ne bɛ maakɔrɔ wele min kɛ ka na sigi
Ɔwɔ an b'a feere	1.468005	Ɔwɔ an b'a feere	Awɔ an b'a feere
[?]	1.532	A jara fɛ a ka ma juguw sɔrɔ	A ka ca ala fɛ a kɛ man di k'o sɔrɔ
A bɛ se ka kɛ an bɛ sen fana lajɛ, an bɛ la sen lajɛ.	2.875986	A bɛ se ka kɛ an bɛ sen fana lajɛ an bɛ sen lajɛ	A bɛ se ka kɛ an bɛ sen fana lajɛ an bɛ na sen lajɛ
Hakɛto bɛ kuma na hali bawoli yɛrɛ.	2.94	Hakɛtobi tumaumana na hali baw wele yɛrɛ	Hakɛto bɛ kuma na hali bawo yɛrɛ
I yɛrɛ b'a fɔ ko [cs] ko fɛn wɛrɛ tɛ nin na ?	2.236009	I yɛrɛ b'a fɔ ko ese ko fɛ wɛrɛ tɛ nin na	I yɛrɛ b'a fɔ ko[cs] ko fɛn wɛrɛ tɛ nin na
I b'a ye ko gɛlɛya b'a ko kɔnɔ kosɛbɛ kosɛbɛ.	2.971995	I b'a ye ko gɛlɛya b'a ku tɛmɛ ku kosɛbɛba kosɛbɛ so	I b'a ye ko gɛlɛya b'a ko kɔnɔ kosɛbɛ kosɛbɛ
Dow ta bɛ taa [?]	1.244	Don ta bɛ taa	Dɔw ta bɛ taa
Sisan aw bɛ jɛgɛ,kɔlɔlɔ min b'a la sisan	2.235986	Sisan aw bɛ jɔ kɔlɔn min' la sisan	Sisan an bi jɛgɛ kɔlɔlɔ min b'a la sisan
Cɛn yɛrɛ kɔni ka fɔ [?]	1.148005	Cɛ yɛrɛ ŋ ka fɔ	Cɛ yɛrɛ kɔni ka fɔ
[um] ne nana daŋaniya sɔrɔ	1.468005	Ɔ ne nana dangaani y'a sɔrɔ	[um] ne nana danganiya sɔrɔ
[cs] sutura don [cs]a bi sutura	2.331995	Bon sutura don parce que a bɛ sutura	[cs] sutura don paseke a bi sutura
Ni ale dayɛlɛ la	1.307982	Ni ale daɛlɛ la	Ni ale dayɛlɛ la
A dɔgɔyalen ka kɛ siɲɛ saba ye kile kɔnɔ an balimamuso a y'a jateminɛ ni dɔw fana ye kɔnɔ ta.	6.62	A dɔgɔlen ka kɛ tiɲɛ saba ye tile kɔnɔ an balimamuso a y'a jateminɛ ni dɔw fana ye kɔnɔ ta	A dɔgɔyalen ka kɛ siɲɛ saba ye kile kɔnɔ an balimamuso a y'a jateminɛ ni dɔw fana ye kɔnɔ ta
K'a sababu kɛ[cs] wagati dɔ kɛra.	1.947982	Ka sababu pa waga de dɔ kɛra	K'a sababu kɛ[cs] wagati dɔ kɛra
Nin lajɛliw de bɛ kɛ pezeli waati musow ka se ka pezeli kɛ nin ye fɔlɔ fɔlɔ.	4.828005	Ni lajɛliw de bɛ kɛ pezeli waati musow ka se ka pezelili kɛ nin ye fɔlɔ fɔlɔ	Nin lajɛliw de bɛ kɛ[cs] waati musow ka se ka pezeli kɛ nin fɔlɔ fɔlɔ
U y'a bila dɛ bawo ni maa t'i bolo i ti se ka fini sɛnɛ, fini fan bɛɛ ye maa ko ye.	4.348005	U y'a bila dɛ bɔ ni maacibolili se ka fini sɛnɛ fini fan bɛ maa koyi	U y'a bila dɛ bawo ni maa t'i bolo i ti se ka fini sɛnɛ fini fan bɛɛ ye maa ko ye
Yaasa i ka se k'i ka bagan nunnu feere.	2.748005	Yaasa e ka segin ka bagan ninnu feere	Yaasa i ka se k'i ka bagan ninnu feere
Kungoda la ka lakanani kɛ u ka jida la, ka lakanani kɛ hali u ka sumanko ninnu minnu bɛ kɛ sababu ye ka tiɲɛni kɛ	7.964014	Duguda la ka lak kɛ u ka jida la ka lak kɛ hali u ka sumanko ninnu minnu bɛ kɛ sababu ye ka tiɲɛni kɛ	Kungoda la ka lakanali kɛ u ka jida la ka lakanali kɛ hali u ka sumanko ninnu minnu bɛ kɛ sababu ye ka tiɲɛni kɛ
O kɔni de tun b'an bolo a wagati la.	2.651995	Aw kɔni de tun b'an bolo a wari la	O kɔni de tun b'an bolo a wagali la[um]
[cs] tɔgɔ de bɛ d' a kan.	1.564	Kɔmuso tɔgɔ de bɛ d' a kan	[cs] tɔgɔ de bɛ d' a kan
O de bɛ taga balan yɔrɔ dɔ la [cs] joli mana na da o bɛ taa ka segin min kɛ ka na da dusukun na	4.603991	O de bɛ taa ba yɔrɔ dɔ la dɔn joli min mina ka taa u bɛ taa ka segin nɛgɛ ka na dususufu na	O de bɛ taga balan yɔrɔ dɔ la[cs] joli min kana ka taa o bɛ taa ka segin min kɛ ka na dusukun na
O kɔfɛ saya bɛ se ka bɔ a la.	1.948	O kɔfɛ saya bɛ se ka bɔ a la	O kɔfɛ saya bɛ se ka bɔ a la
Ka sumaya da u la.	1.372	Ka suma da aw la	Ka sumaya da o la
A minɛ dafalen ka sɔrɔ i la [?]	1.851995	A minɛ dafalen ka sɔrɔ i na	A minɛn dafalen ka sɔrɔ i la[?]
Ɔ baara kɛ waatiw kɔnɔ[?]	1.276	O baara kɛ waatiw kɔnɔ	O baara kɛ waatiw kɔnɔ
An b'o [cs]ninnu de walanka.	2.011995	An b'o si fɔ ninnu de walilanga	An b'o sifɔn ninnu de wala nka
An molu yafa de.	1.147982	An malo yakarire	An m'olu yakakari
[um] la [um]	1.531995	Ɛɛ la ɛɛ	[um] la[um]
N'i ye mɔgɔ kunbɛ cogo min na fana o de bɛ	2.364014	N'i ye mɔgɔ kun bɛ cogo min na fana o de bɛ yen	N'i ye mɔgɔ kunbɛ cogo min na fana o de bi[?]
O tulonkɛ ninnu tun ye maloya de y'an bolo.	1.724	O tulonkɛ ninnu tun ye maloya de anan bolo	O kulonkɛ ninnu tun ye maloya de ye an bolo
Ni ka se ka dan kari kɛrɛfɛdɛn in kɛlɛli la.	3.388005	Ni ka se ka dan ka di kɛrɛfɛden kɛlɛ la	Ni ka se ka dankari kɛrɛfɛdɛn kɛlɛli la
A dafalikɛ o majɛ nunnu na.	1.692	A dafalikɛ o majɛ ninnu na	A dafalikɛ o manje nunnu na
An ga wari kana jasi an ga se ga an [pause]an me munumunu na fɛn min kama yɛrɛ	4.54	An wari kana jate an ka se ka an bɛ munumunuuna fɛn min kama yɛrɛ	An wari kana jate an ka se ka an bɛ munumunu na fɛn min kama yɛrɛ
Yarɔ o fana ma nɔgɔn dɛ.	1.02	Ya'rɔ fana ma nɔgɔɔgɔn dɛ	Yarɔ o fana ma nɔgɔn dɛ
[?] nate pase ba tigɛ ye fɛn dɔ ye a sufɛlama ma ɲi.	3.291995	Atɛ pa que ba tigɛ ye fɛn dɔ ye a sufɛlalama ma ɲi	A na te pasiki batigi ye fɛn dɔ ye a sufɛlama in
[?] o b'a yira k'a fɔ [?]	1.308005	O b'a jira k'a fɔ	[cs] o b'a yira k'a fɔ
Bɛnkan de don n'a bɛnn'a la an b'a kɛ sarakokelen ye	2.524014	Bɛnkan de don n bɛndala ala kɛ sira o kelen	Bɛnkan de don n'a bɛnn'a la an b'a kɛ sirako kelen ye
N'u nana wakati min na a bɛ sɔrɔ ka n'a lajɛ.	2.075986	N'u nana waati min na a bɛ sɔrɔ ka lajɛ	N'u nana wagati min na a bi sɔrɔ ka n'a lajɛ
Ni naaninan digina.	1.116009	N nana digina	Ni naaninan jiginna
Ola an b'a dɔn ko ɔ nin bana suguya de b'a la.	2.908005	Nka an b'a dɔn ko ɔ nin kana sidiya de b'a la	O la an b'a dɔn ko[um] nin bana suguya de b'a la
Bɔn an bɛna a kun cɛ ni mun ye	1.436	Ban an bɛna k'u cɛ ni mun ye	[cs] an bɛna kun cɛ ni mun ye
[cs] n'an ye fura k'u la .	1.308005	Don ye fura k'u la	[cs] n'an ye fura k'u la
N'olu bɛ baarafin kɛ	1.756009	N'olu de bɛ baarafin kɛ	N'olu bɛ baarafin kɛ
Kala den hakɛ mun y'a bila.	1.436009	Ka ni hakɛmuya bila	Arajo hakɛ mun y'a bila
I y'a ye [?] [cs] a ye n'a fɔ sisan [?]	1.788005	O y'a yan bon a ye naa fɔ sisan	I y'a ye a[cs] a ye n'a fɔ sisan
[?] nin [?]	1.372	A ni ma	Aa nin kɔnɔna na
A bi a bi da dugu de jukɔrɔ	1.403991	A a bɛ dugu de cogokɔrɔ	A bi a bi da dugu de jukɔrɔ
Anw de b'a lajɛ a be ɲ'an bolo cogo min na an kɔni be k'a kɛ ten [cs]mɛ k'a fɔ ko gɛlɛya wɛrɛ be an ni dugumɔgɔw cɛ	5.948005	Anw de b'a lajɛ a bɛ nin y an bolo cogo min na an kɔniɔni bɛ k'a kɛ tɛmɛ k'a fɔ ko gɛlɛya wɛrɛ bɛ an ni dugumɔgɔw cɛ	Anw de b'a lajɛ a bɛ ɲɛ an bolo cogo min na an kɔni bɛ k'a kɛ ten mɛ k'a fɔ ko gɛlɛya wɛrɛ bɛ an ni dugumɔgɔw cɛ
Bu nun fɛnɛ ye o bana de ye [?]	1.275986	Busɛn o banadɔ ye	Bulu fɛnɛ ye o bana dɔ ye
Me se ka ca caman yɛrɛ de kɛ n'a wari ye.	2.14	Ne tɛ ka ca caman yɛrɛ kana n'a wari ye	N bɛ se ka caman caman yɛrɛ de kɛ n'a wari ye
[um] u ka ko o togow laten [?]	1.5	Ɔ u ka ko ougu ladi	[um] u ka ko o cogo la ten
An bɛ taabolo min kan nin ye fo ka taa diɲɛ kɛ min ye nansara tɛ kun an na duguni koyi	3.356	An bɛ taabolo min ye f fo ka da dunkɛ min na tɛ k anna duguin ko ye	An bɛ taa bolo min kan nin fɔ ka taa diɲɛ min nanzara tɛ kum'a la tugunni koyi

End of preview. Expand in Data Studio

📘 African Next Voices – Bambara (AfVoices)

The AfVoices dataset is the largest open corpus of spontaneous Bambara speech at its release in late 2025. It contains 423 hours of segmented audio and 612 hours of original raw recordings collected across southern Mali. Speech was recorded in natural, conversational settings and annotated using a semi-automated transcription pipeline combining ASR pre-labels and human corrections. We release all the data processing code on GitHub.

🔎 Quick Facts

Category	Value
Total raw hours	612 h (1,777 raw recordings; publicly available on GCS)
Total segmented hours	423 h (874,762 segments)
Speakers	512
Regions	Bamako, Ségou, Sikasso, Bagineda, Bougouni
Avg. segment duration	~2 seconds
Subsets	159 h human-corrected, 212 h model-annotated, 52 h short (<1s)
Age distribution	Broad, across young to elderly speakers (90% between 18 and 45)
Topics	Health, agriculture, Miscellaneous (art, education, history etc.)
SNR distribution (raw recordings)	71.75% High or Very High SNR
Train / Test split	155 h / 4 h

Motivation

The African Next Voices (ANV) project is a multi-country effort aiming to gather over 9,000 hours of speech across 18 African languages. Its goal is to build high-quality datasets that empower local communities, support inclusive AI research, and provide strong foundations for ASR in underrepresented languages.

As part of this initiative, RobotsMali led the Bambara data collection for Mali. This dataset reflects RobotsMali’s broader mission to advance AI and NLP research malian languages, with a long-term focus on improving education, access, and technology across Mali and the wider Manding linguistic region.

🎙️ Characteristics of the Dataset

Data Collection

Speech was collected through trained facilitators who guided participants, ensured audio quality, and encouraged natural, topic-focused conversations.
All recordings are spontaneous speech, not read text.
A custom Flutter mobile app (open-source) was used to simplify the process and reduce training time.
Geographic focus: Southern Mali, to limit extreme accent variation and build a clean baseline corpus.

Segmentation and Preprocessing

Raw audio was segmented using Silero VAD, retaining ~70% of the original duration.
Segments range from 240 ms to 30 s.
Voice activity detection helped remove long silences and improve data usability.

Transcriptions

Pre-transcribed using the ASR model soloni-114m-tdt-ctc-v0.
Human annotators corrected the transcripts.
A second model (soloni-114m-tdt-ctc-v2) was trained using the corrected transcripts and used to regenerate improved labels.
Two automatic transcription variants exist for each sample: v1 (from soloni-v0) and v2 (from soloni-v2).

Acoustic Event Tags

The following tags appear in transcriptions:

Tag	Meaning
`[um]`	Vocalized pauses, filler sounds
`[cs]`	Code-switched or foreign word
`[noise]`	Background noise (applause, coughing, children, etc.)
`[?]`	Inaudible or overlapped speech
`[pause]`	Long silence (>5 seconds or >3 seconds at segment boundaries); due to VAD segmentation this tag is rarely used

📂 Subsets

1. Human-corrected (159 h, 260k samples)

Fully reviewed and corrected by annotators.
Only subset with a definitive text field containing the validated transcription.

2. Model-annotated (212 h, 355k samples)

Includes automatic labels: v1 (soloni-v0) and v2 (soloni-v2).
No human review.

3. Short subset (52 h, 259k samples)

Segments <1 second (formulaic expressions, discourse markers).
Excluded from human annotation for optimization purposes.
Automatically labeled (v1 & v2).

⚠️ Limitations

Clean dataset vs real-world noise: Over 70% of recordings can be categorized as relatively clean speech. Models trained solely on this dataset may underperform in noisy street or radio environments typical in Mali. See this report if you are interested in learning more about the strengths and weaknesses of RobotsMali's ASR models.
Reduced code-switching: French terms were often replaced by [cs] or normalized into Bambara phonology. This improves model stability but reduces realism for natural bilingual speech.
Geographic homogeneity: Focused on the southern region to control accent variability. Broader dialectal coverage might require additional data.
Simplified linguistic conditions: Overlaps, multi-speaker settings, and conversational chaos are minimized—again improving training stability at the cost of deployment realism.

📑 Citation

@misc{diarra2025dealinghardfactslowresource,
      title={Dealing with the Hard Facts of Low-Resource African NLP}, 
      author={Yacouba Diarra and Nouhoum Souleymane Coulibaly and Panga Azazia Kamaté and Madani Amadou Tall and Emmanuel Élisé Koné and Aymane Dembélé and Michael Leventhal},
      year={2025},
      eprint={2511.18557},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2511.18557}, 
}

You may want to download the original 612 hours dataset with its associated metadata for research purposes or to create a derivative. You will find the codes and manifest files to download those files from Google Cloud Storage in this repository: RobotsMali-AI/afvoices. Do not hesitate to open an issue for Help or suggestions 🤗

Downloads last month: 227

Size of downloaded dataset files:

142 GB

Size of the auto-converted Parquet files:

142 GB

Number of rows:

874,762

Models trained or fine-tuned on RobotsMali/afvoices